15 results about "Topological information" patented technology

The invention discloses a manifold perception-based large model navigation method guided by a historical topological graph, and relates to a computer vision technology. The method aims at solving the challenges that in the navigation process, long-distance reasoning experience is insufficient, instruction fragments and dynamic visual observation are difficult to align, and large model reasoning is prone to illusion interference. Firstly, a large model based on an encoder-decoder structure is used for supplementing historical information coding for a visual observation sequence, and therefore global topological information guidance is provided for long-distance reasoning. And secondly, in order to effectively solve the problem that large model reasoning is subjected to illusion interference, significant space-time differences in a visual observation sequence are mined by using a multi-curvature manifold, so that the large model can accurately describe the current environment and make a decision according to a visual reference object in thinking. Besides, in order to strengthen the perception capability of the large model to the space structure and establish a graph self-attention mechanism, the node distance embedded in the constructed historical topological graph is combined with the visual similarity so as to model the space relationship between the nodes.

The invention discloses a natural resource information survey management method and system and a storage medium, and relates to the technical field of natural resource information survey management, and the method comprises the steps: obtaining the patrol information of a survey region, carrying out the vectorization of the map information of the survey region, and obtaining the simplified topological information of the survey region; performing spatial information extraction on the map information of the investigation area to obtain hierarchical distribution information of the investigation area; and according to the planning information of the investigation area, pre-planning information of the investigation area is preset. According to the method, the survey area is vectorized in advance to obtain the contour, the processing requirements of data loading and storage and the like are reduced, and the communication difficulty of a field survey use scene is reduced. And meanwhile, by simplifying the topology, the use intensity in an off-line scene can be reduced, and the use requirements on various scenes of natural resource investigation can be well met. According to the invention, the disadvantage in the prior art that a natural resource management scheme cannot comprehensively consider the use environment scene and the off-line resource for overall planning is effectively solved.

The invention provides a method and a system for quantifying a forest crown space structure of a sand-fixing forest based on a foundation and an unmanned aerial vehicle laser radar, and is applied to the technical field of data processing. According to the method, the fixed sarin quantized multi-source core data is obtained, and the standardized fusion point cloud data set is generated through point cloud preprocessing and registration correction. The method comprises the following steps: performing individual tree and forest stand scale layering on a data set, constructing ternary association nodes, aggregating topological information, and generating a forest crown structure feature association graph; then, core features are extracted through neural network modeling, high-dimensional feature embedded vectors are generated, a quantitative basic model is constructed through combination of hierarchical quantization engine iterative optimization and actual measurement calibration, finally, precision verification and dynamic correction are completed through combination of actual measurement and growth monitoring data, and multi-index requirements are balanced; and outputting a forest crown space structure quantification result adaptive to the Sarmonin full-monitoring scene.

The invention discloses an image generation method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring map topology information, traffic flow simulation configuration data and user semantic description information; and generating an automatic driving model picture corresponding to the user semantic description information, the traffic flow simulation configuration data and the map topology information according to a large language model. According to the technical scheme provided by the embodiment of the invention, by acquiring the map topology information, the traffic flow simulation configuration data and the user semantic description information, the geometric position, the scale and the road structure logic consistency of the traffic participants can be maintained; road images of various special scenes can be generated by generating automatic driving model pictures corresponding to the user semantic description information, the traffic flow simulation configuration data and the map topology information according to a large language model; the problems that in the prior art, automatic driving long-tail scene data is scarce, generation controllability is insufficient, and semantic consistency is poor are solved.

The present application relates to the field of satellite computing power network task scheduling, and provides a satellite computing power network-oriented dependent task scheduling method, device and readable storage medium, the satellite computing power network includes a plurality of interconnected low earth orbit satellite nodes, the task scheduling method includes the following steps: S100: receive the original adjacency matrix of DAG task and the initial feature vector of subtask, perform topological symmetry, extended neighborhood definition and structure position coding, and construct a bidirectional information flow embedding space;S200: based on the symmetric adjacency matrix and the initial feature of the subtask, the DAG task is processed through the graph neural network containing multi-head attention and GRU gate, and the advanced embedding feature containing global topological information is extracted;S300: based on the extracted advanced embedding feature, a state space, an action space and a reward mechanism are constructed to drive the agent to generate a scheduling strategy. Therefore, the joint scheduling for the global topology can be realized, and the satellite environment of dynamic and resource-limited can be better adapted.

The application discloses a human action prediction method and device based on action-aware dynamic graph structure, and belongs to the technical field of computer vision and graph neural network. By acquiring a historical motion sequence, and performing mask joint reconstruction on the sequence, a dynamic attention graph for representing spatial dependency between joints under a current action mode is generated. The dynamic attention graph is fused with a shared adjacency matrix to construct an action-aware adjacency matrix. Based on the action-aware adjacency matrix, graph convolution is performed on the historical motion sequence to extract features rich in spatial topological information. Multi-scale time sequence convolution is performed on the spatial features to capture local details and long-range dependencies under different receptive fields, and time sequence features are extracted. Future motion sequences are predicted according to the time sequence features. The application realizes adaptive spatial modeling and fine-grained time sequence modeling of action modes, and improves the accuracy and naturalness of human action prediction.

The application relates to a kind of wisdom water affairs real-time monitoring early warning method and system based on Internet of Things, solve the problem that abnormal identification, mechanism simulation and decision link are disconnected in the prior art, leading to traceability difficulty, early warning delay, its method includes: through edge node real-time collection analysis water quality data and identify abnormality, aggregate data after uploading to cloud and simulate to generate global risk atlas reflecting abnormal distribution and diffusion trend.Accordingly, dynamically enhance the monitoring of key areas, lock the suspected source in combination with topological information, and finally comprehensively evaluate the hazard level and generate a graded early warning push.The application has the following effects: through edge-cloud collaboration and mechanism data fusion, a closed-loop system from abnormal perception, dynamic tracing to intelligent decision-making is constructed, and the ability of water affairs risk from passive alarm to active prediction is upgraded.

The application provides a method and system for quantifying the spatial structure of sand-fixing forest canopy based on ground and unmanned aerial vehicle laser radar, and is applied to the technical field of data processing. The application obtains quantified multi-source core data of sand-fixing forest, generates a standardized fusion point cloud data set through point cloud preprocessing and registration correction. The data set is layered at a single tree and stand scale, a three-element correlation node is constructed, and topological information is aggregated to generate a canopy structure feature correlation graph. Then, a neural network model is relied on to extract core features, generate a high-dimensional feature embedding vector, and iterate and optimize through a layered quantification engine combined with actual measurement calibration to construct a quantification basic model. Finally, the accuracy verification and dynamic correction are completed in combination with the measured and growth monitoring data, the multi-index requirements are balanced, and the canopy spatial structure quantification results suitable for the whole monitoring scene of sand-fixing forest are output.

The application discloses a scene matching method combining semantic information and topological features of ground objects, and belongs to the technical field of geographic information processing. The method comprises the following steps: dividing a city into a street block scene set by taking a main road as a boundary; calculating a scene centroid by using the center and area attributes of each building monomer in the city scene; establishing a polar coordinate system and calculating the coordinates of each building monomer in the city scene; dividing the polar coordinate system into sub-regions, counting the number of building monomers in each sub-region in the sub-region set, and obtaining a statistical histogram about the whole scene; obtaining a set of scenes to be matched, calculating the statistical histogram about the building monomers for each scene to be matched, calculating the similarity between the scene to be matched and any scene according to the difference degree of the statistical histogram, finding the best matching scene, and obtaining a matching scene set of the scene to be matched. The application can fully utilize the semantic structure information and spatial topological information of ground objects, and provides technical support for local scene interpretation and matching in a city and other applications.

A point cloud classification and segmentation method based on geometric perception and parallel attention is proposed, comprising the following steps: 1) preprocessing the point cloud classification and segmentation dataset; 2) constructing a geometric perception module; 3) designing a parallel attention module, dividing it into efficient channel attention and spatial attention branches, and modeling feature channel weights and local spatial dependencies respectively; 4) building an end-to-end network structure based on geometric perception and parallel attention, and using the training set to complete model training and parameter optimization; 5) loading the trained network parameters, performing end-to-end inference on the test set, and outputting point cloud classification results and point-by-point segmentation results. This invention fully mines local spatial topological information, enhances the model's ability to represent local features of complex-shaped targets, and effectively reduces computational overhead. It aims to solve the core problems of current methods, such as insufficient ability to extract local geometric information, inaccurate mining of key feature channels and spatial information, and high redundant computational overhead.

The invention relates to the technical field of artificial intelligence, in particular to a knowledge graph embedding method and system based on semantic and structure unified modeling, and the method comprises the steps: generating entity enhanced representation through fusing semantic and local topological information in a knowledge graph; performing complex number transformation on the entity enhancement representation, and dynamically adapting to the structural context of the knowledge graph based on an attention mechanism to obtain a final transformation representation; candidate negative samples are dynamically generated in a self-adaptive mode according to the relation attributes, hard negative samples are mined, and negative samples are obtained by randomly sampling the candidate negative samples and splicing the candidate negative samples with the hard negative samples; and setting a loss function according to the negative sample scale and a preset loss type, and performing comparative learning on the knowledge graph embedding model by using positive and negative samples based on the loss function. Through unified modeling of semantic information and graph structure embedding of entities and relationships in the knowledge graph, the method can be suitable for application of knowledge graph completion in scenes such as intelligent questions and answers and service recommendation.

This invention provides a method and system for main transformer load forecasting based on power big data. The method includes: determining target feeders based on the power supply range of the target main transformer and the affiliation of each feeder; determining the feeder-associated area based on the geographical and spatial topological information of the power supply area of ​​each target feeder; determining the initial load contribution based on the load time-series data of the power supply area corresponding to each feeder in the target feeder pair and the load change trend of each feeder within the feeder-associated area; performing load synchronization analysis based on the load feedback data of each feeder and each feeder to be matched to obtain a superimposed effect feeder combination; adjusting the initial load contribution of each feeder based on the load fluctuation correlation data between feeders within the superimposed effect feeder combination to obtain the preliminary load forecast results for each superimposed effect feeder combination, and then superimposing them to obtain the final load forecast result. This invention improves the rationality of power dispatching decisions and the stability of power supply.

A point cloud segmentation method based on Gaussian semantic sampling and multi-scale feature fusion is proposed, comprising the following steps: 1) preprocessing the dataset; 2) designing an adaptive Gaussian semantic sampling algorithm to perform AGSS sampling on the input projection features to obtain a more representative point set; 3) proposing a multi-scale fusion VLAD module to process the features extracted by each layer of the encoder in parallel, constructing a global descriptor with rich details and strong semantic discriminativeness; 4) loading the trained network parameters, performing end-to-end inference on the test set, and outputting the point cloud segmentation results. This invention, through the adaptive Gaussian semantic sampling algorithm and the multi-scale fusion VLAD module, fully mines local spatial topological information, enhances the model's ability to represent local features of complex-shaped targets, and effectively reduces computational overhead through lightweight design. This solves the core problems of current methods, such as insufficient extraction of local geometric information, inaccurate mining of key feature channels and spatial information, and high redundant computational overhead.

The invention provides an online switching method of parallel training strategies and a deep learning model training system. The method comprises the following steps: determining a fragmentation mode of new communication group information and model parameters according to new parallel dimension division information and new working node topological information, and further generating a mapping scheme for fragmentation migration of the model parameters according to old parallel dimension division information and old working node topological information; and migrating the model parameters to the new working node based on the fragmentation mode and the mapping scheme. The new communication group information may be used to enable a communication group, and the enabled communication group and the model parameters migrated to the new working node are used for the new working node to perform parallel model training under the new parallel dimension division. Therefore, according to the method and the device disclosed by the invention, the corresponding communication group is started during operation, the parameter mapping is reconstructed, and the re-fragmentation and migration with the minimum cost are executed, so that the online rapid switching of the parallel training strategy is realized.

The invention provides an automatic measurement path planning method, device and equipment for a rotary body. The automatic measurement path planning method comprises the following steps: uniformly converting an obtained to-be-measured rotary body into a Brep model containing geometric information and topological information; determining topological information related to a target to-be-detected surface according to the target to-be-detected surface selected by the user; determining a corresponding measurement strategy according to the measurement type to which the target surface to be measured belongs, and generating a three-dimensional coordinate value for measuring a measurement point of the target surface to be measured according to the measurement parameter and the measurement strategy input by a user; based on the constructed measuring head used for detecting the motion trail of the workpiece, according to the target to-be-measured surface, the measuring strategy and the three-dimensional coordinate value, a measuring path of the measuring head for measuring the target to-be-measured surface is generated according to the principle of no collision and the shortest measuring path, and the rotating body to be measured is measured according to the measuring path. Therefore, not only can the measurement quality and efficiency of the rotary body part be improved, but also the product assembly precision can be guaranteed.